Keyboard calligraphy

May 1, 2008 12:25 AM

Keyboard calligraphy"To produce such a typeface, Müteferrika knew he had to analyze Arabic script. Calligraphers might learn to make the correctly shaped letter combinations by practice, without conscious application of tens of thousands of rules, but for machine reproduction of the script, deciphering those rules was exactly what was essential."
posted by dhruva (28 comments total)

I've always been curious why western scripted computer fonts have a glyph for each letter instead of a glyph for each join of two letters split at the centerline of each character. For example the word 'Them' could be represented by four glyphs T h e & m, or five glyphs capital T start, Th join, he join, em join, and m end. Admittedly you'd need from 725-810 glyphs to represent English script (more for 3 character ligatures), but letters would automatically be kerned, and two character ligatures would be built in. As well the end of each letter would seamlessly merge with the next at the appropriate height.

I'm sure someone who knows something about typography and design could explain to me why such a system is a bad idea.
posted by BrotherCaine at 12:55 AM on May 1, 2008

Arabic is such a beautiful and crazy system of writing.
posted by BrotherCaine at 12:58 AM on May 1, 2008

Arabic is such a beautiful, crazy and peace-loving system of writing whom I have the utmost respect for (there is no other God but Allah (peace be on him)) .
posted by basicchannel at 1:02 AM on May 1, 2008

The complex patterns, flowing linguistic system, a blossom of script.
posted by mrloiq at 1:09 AM on May 1, 2008

"...result is that the individual letters in a well-written piece of text are in constant motion, like dancers in a polonaise: In the course of the dance, they bow to each other, embrace each other, push each other away, hug each other’s necks and fall at each other’s feet..."

I hate Arabic calligraphy for exactly this reason. For native readers, who've grown up deciphering its abstruse shenanigans, it's considered rather beautiful. But for those of us who grew up reading printed books and Sans Serif signs and inscriptions, Arabic calligraphy is something we'll never get used to.

I think Mulder's Saudi Aramco piece makes rather too much of the calligraphy part though. Arabic handwriting style (as opposed to decorative calligraphy) was fairly amenable to movable type printing-three glyphs per character depending on position seems to be all there really was to it. I don't see where this "tens of thousands of rules" malarkey comes from.
posted by jackbrown at 1:13 AM on May 1, 2008 [1 favorite]

err...four glyphs per character. I forgot about the occasional free-standing letter.
جال براون
posted by jackbrown at 1:17 AM on May 1, 2008 [1 favorite]

Arabic typeface? Wow, that was one crazy Müteferrika!
posted by flapjax at midnite at 1:22 AM on May 1, 2008 [1 favorite]

But nice to see this on MetaFiruta!
posted by flapjax at midnite at 1:23 AM on May 1, 2008

Great post, thank you very much.

I've always found something entrancing about Arabic Calligraphy. On my first visit to Turkey last year I came across some lovely modern artwork incorporating it by an artist called Süleyman Saim Tekcan - his etchings in particular are stunning.
posted by protorp at 2:38 AM on May 1, 2008

BrotherCaine, what you're talking about there is a "stroke font", or "stroke-based font", and they are used extensively for Asian languages, and also for CAD of various kinds including cutting.
posted by aeschenkarnos at 6:19 AM on May 1, 2008

BrotherCaine, such systems, called syllabic writing, have existed for a long time.
Writing, as a job, started off as religious record keeping, and existed in the form of pictographs or logograms. Pictures for words, or, in the case of logograms, the ideas behind words. Egyptian heiroglyphs and such.

Things progressed from there. Writing was adopted as a means of civic recordkeeping; mainly tracking allotments of food from granaries, and then, over time, tracking trade between cultures. It was during this time that writing evolved from pictographs into syllabics.

One problem (as you noted) is the massive number of glyphs you'd need. You can get a sense of how big this problem would be by looking at the syllabary for Katakana. Combining five english consonants with the english vowels would require twenty-five syllabic glyphs. I don't think there are many complete syllabaries with all possible syllabic combinations in them, if there are any at all.

Another problem is that spoken words don't always fit neatly into a syllabic system. You can say "squirrel", but your writing may not have a glyph for that syllable exactly. So it gets munged up a bit, perhaps into two gylphs, like (squi) (ral) or maybe just (squi), with the contextual understanding that there's a bit of the word left off for convenience's sake.

Alphabets are a refinement of syllabic systems; they address the issues of ambiguity and complexity that were implicit in those earlier systems.
posted by boo_radley at 6:56 AM on May 1, 2008

Fascinating stuff—thanks, dhruva!
posted by languagehat at 7:02 AM on May 1, 2008

I'm sure someone who knows something about typography and design could explain to me why such a system is a bad idea.
because you'd need 700+ glyphs instead of 50. Because people are taught how to write letter-by-letter and that translates to -- yes! -- typing letter by letter.

What's the advantage? There are better ways to do automatic kerning and ligatures.
posted by bonaldi at 7:59 AM on May 1, 2008

I've always been curious why western scripted computer fonts have a glyph for each letter instead of a glyph for each join of two letters split at the centerline of each character.

Because it's considered better to have a system that only stores the exceptions rather than to store everything. There aren't many ligatures, so storing everything as a ligature is a waste. And kerning is just a number for each character pair... you don't need a set of glyphs for every one.
posted by smackfu at 8:10 AM on May 1, 2008

File under: Things I've Never Thought About Until Now But Are In Reality Effing Fascinating.
posted by redsparkler at 11:22 AM on May 1, 2008

BrotherCaine, who is not talking about stroke fonts or syllabaries, raises an interesting question. I'm gonna put on my Marxist hat and say the answer is historical necessity.

In the original home of alphabetic movable metal type, Gutenberg's shop, they made separate letterpunches for every letter, as well as for dozens of ligatures. This was the logical thing to do: The goal was to efficiently make texts like the existing (handwritten) ones, so he imagined a text handwritten, then divided it into glyphs at places where a straight vertical line of white lay. The imitation was very effective—when this massive quantity of Bibles appeared on the market, Gutenberg was accused not of illicit printing, but of using demonic aid to speed manuscript copying. Having all those ligatures helped, but what helped as much was that the standard handwriting of the day was a straight-sided blackletter that didn't need much kerning. If they'd been using the humanist letterforms we're used to, or Arabic script, or anything else that requires lots of individual adjustments to the glyphs, the printed nature of the thing would have been obvious; he would have lacked the time, money, space, kerning technique, and organization to make and manage all those hundreds of variant letterpunches.

Fast forward to the typewriter era. Printers have got kerning figured out by this point. But more complicated adjustments (like for Arabic) are still beyond printing, the typewriter can't even kern, and most importantly, the glyph-management problem is still there: Limited space for a keyboard, limited length for a type bar. Got to cut down on the number of symbols.

Also at this point Hollerith is developing the punch card. His goal is to represent the information content of writing, not the typography, and he's got to worry about all kinds of data density problems. But of course that means he also assigns one code to each letter.

We'll fast forward to ASCII in a moment, but let's take a side trip to the Linotype. Elegant printing it wasn't, but boy was it fast, and that is the tradeoff that newspapers have always made. (That's why styles of book publishers employ italics, and styles of papers don't: Italics are tricky, and papers couldn't waste time handling them.) I will spare you my extensive rhapsodies over the utter brilliance that is the Linotype, but suffice it to say that it vastly eased the management of all those little bits of metal, automated the justification of text, and produced slugs suitable for printing with a repertoire of only 90 glyphs. That's upper case, lower case, ligatures, numerals, punctuation, the whole shebang. In its later years it even converged with a descendant of Hollerith's invention: The TeleTypeSetter could cast slugs as instructed by a paper tape.

So those teletypewriter codes were sort of multiplying, and becoming inadequate, and so in 1960 a bunch of Americans sat down to Standardize a Code for Information Interchange. Our friend ASCII has 94 printing characters, space, and 33 control codes, for a total of 128 codes (just fitting in seven bits). Abundant for giving a code to each letter, worthless for giving a code to each pair of letters.

Not really quite sufficient for non-American typography with accented characters, either. So lots and lots and lots and lots of incompatible eight-bit supersets of ASCII were invented to cover other languages, until finally the Unicode people and the ISO both sat down and decided we needed an encoding with room for all the characters. This was possible because storage and bandwidth and processing costs had finally dropped to the point where we could give a character a 16-bit code. You'd think 65,536 codes would be enough, wouldn't you? (Nope, they weren't. Thence all the talk of the Basic Multilingual Plane versus surrogates, and thence also tragicomedies like Han unification. Anyway.)

Meanwhile, computers were getting better and better at the whole font business. In my childhood we had a printer with a built-in font (one size, bold and italic, monospaced): The computer told the printer what the letter was, a la Hollerith, and the printer knew how to lay down the inkspots for it. Then there were bitmap fonts in the computer: One little picture for each letter, digit, or mark, and the printer would receive them from the computer. Eventually printers got good enough that they could keep everything in alignment all the time, a precondition for making BrotherCaine's half-letters match up.

Which is to say, we're now in a position for BrotherCaine's approach: We have the physical technology and we can manage the information for all those letter pairs. But as bonaldi and smackfu point out, there's no need. Because while all of this was being invented, we were also coming up with outline fonts and good computer typesetting. Outline font technology (PostScript, Adobe Type 1, Type 3, TrueType, OpenType sound familiar?) took the approach that the font isn't pictures, but drawing instructions, so the letters could be drawn at any size and resolution. Modern niceties like hinting were developed to refine those size changes: The glyph for the 8-point letter is not just the glyph for the 16-point letter at half size. In the 1970s Don Knuth got tired of the lame computer typesetting available then and developed TeX, which modeled and automated all of the different visual and aesthetic principles that a competent human typesetter has to balance. (And yes, they are numerous. I don't know about thousands, but when you consider glyph-specific things like kerning or parenthesis massage, hundreds is a reasonable guess.) Nowadays technologies like AAT and Graphite attach that aesthetic code to the font, so that a standard rendering engine can make really good typography given a font and a sequence of plain old, non-ligature Unicode characters.

So to summarize, the split-letter approach has never been done because the conditions sufficient for it were also (or were closely followed by) the conditions that made it unnecessary, whereas the whole-letter approach was favored first by necessity and then by sufficiency from the start.

That was fun! Too bad it did nothing for getting my research report written...
posted by eritain at 12:31 PM on May 1, 2008 [9 favorites]

[ffi] [fi] [fl] [ff]

I'm always amazed at how rarely these come out of the case.
posted by Sys Rq at 12:57 PM on May 1, 2008 [1 favorite]

usually only when affluent finance officials have been floating in a fjord, I find
posted by bonaldi at 1:31 PM on May 1, 2008 [3 favorites]

The California Job Case is rather unforgiving when it comes to Norwegian inlets, sadly.
posted by Sys Rq at 1:43 PM on May 1, 2008

California Job Case.

Alas, the reminiscences I linked to have gone bye-bye. Wish I'd quoted more.

*shakes fist at forgetful internet*
posted by languagehat at 2:03 PM on May 1, 2008 [1 favorite]

Bonaldi, please give me a for instance of automatic ligatures. I'd very much like to see it. The system I mention would be some kind of intermediate presentation layer (I don't know jack about typography obviously), presumably characters would still be typed and stored one by one.
posted by BrotherCaine at 5:10 PM on May 1, 2008

Shoot, I realized I was talking at a somewhat abstract level about a field where I don't understand the appropriate jargon to use. The specific problem that it seems like most of the script fonts trying to emulate cursive handwriting I've seen fall down on is where the letters join. From my limited experience with script fonts, the stroke at the end of each character is usually matched up to the next character at the same height. It seems that this is a decision made more for simplifying either the number of characters or the complexity of the font system than for aesthetic reasons. When you look at beautiful cursive handwriting, where the letters join seems to vary for different letter pairs.
posted by BrotherCaine at 5:51 PM on May 1, 2008

BrotherCaine, you'll need an application with automatic ligature support, and an appropriate typeface. For instance, with Minion in Quark Xpress (any version since 3), I can type "affluent finance officials floating in a fjord", and the ffl, fi, ffi, fl and fj will all be replaced with ligatures. It looks lovely.

Most systems don't support this for every letter, because there isn't much to be gained with a lot of character pairs. Those ones are special becaues otherwise they clash horribly. the system you're describing is where a character can be designed to be aware of those around it so they merge nicely. Only Quickdraw GX (that I'm aware of) could do this, and it's a long dead Apple-only thing.
posted by bonaldi at 6:35 PM on May 1, 2008 [1 favorite]

where a character can be designed to be aware of those around it so they merge nicely

The successor to QuickDraw GX is Apple Advanced Typography, and a lovely showpiece for its ligatures is the face Zapfino. The letter joins are very different for different letter pairs. If you've got access to a recent Mac, I strongly advise you to fire it up and type in some worthwhile words; the rendering is delicious.

There are other smartfont architectures for non-Apple platforms; Graphite is one that's open-source, developed by linguists to meet various interesting needs (like different rules for arranging multiple diacritics, depending which language you're in, or rendering discrete tone indicators as a continuous, ligated contour).
posted by eritain at 8:23 PM on May 1, 2008 [2 favorites]

Wow, Zapfino is cream my jeans beautiful. I don't have access to a Mac, but I may install Ubuntu with the Pango-Graphite package just to see what Graphite will do.
posted by BrotherCaine at 9:55 PM on May 1, 2008

Oh, I forgot about ATSUI and all that. I didn't think they had the sort of contextual awareness that lets you put this S after that e and so forth, but I'm wrong. Nice.
posted by bonaldi at 9:30 AM on May 2, 2008

The specific problem that it seems like most of the script fonts trying to emulate cursive handwriting I've seen fall down on is where the letters join.

How would using two character ligatures for every pair solve this problem? Wouldn't you still have iffy joins between the sets of pairs?
posted by smackfu at 1:18 PM on May 2, 2008

smackfu, BrotherCaine is talking about splitting each character into halves. For instance, the word "bat" would be made up of four "hemi-characters" instead of three characters:

1. The first half of the "b" glyph
2. The Joining hemicharacters that form the middle half of the construct "ba"
3. The joining hemichacters that form the middle half of the contstruct "at"
4. And finally, the trailing half of the "t" glyph

To all you bigger typsetting nerds, I apologize if I mixed up "character" and glyph"
posted by Xoder at 6:38 PM on May 2, 2008